Algorithms Of Aeronautical Remote Sensing Image Registration And Stitching

Image splicing refers to a process of splicing a plurality of images obtained from the same scene into a complete image through a specific transformation.With the rapid development of aviation technology,aerial remote sensing images have been widely used in many fields.However,in the process of aerial remote sensing image acquisition,due to the influence of factors such as scattering of sunlight,vibration of the aircraft,and displacement of the image caused by the movement of the aircraft during exposure,the quality of the image obtained is actually poor and the resolution is low;In the aerial photography process,the sequence images acquired by the camera at a low frame rate contain less overlapping areas,and some conventional registration methods cannot be used or cannot obtain high accuracy.Therefore,the splicing of aerial remote sensing images has become a research hotspot in recent years.This paper takes the aerial remote sensing image registration and splicing as the research content and mainly completes the following aspects:1.Aiming at the characteristics of low resolution,few overlapping area,and repeated characteristics of aerial remote sensing images,a two-step registration method with coarse registration and then fine registration was designed.Due to the characteristics of aerial remote sensing images,there are few feature points pairs extracted directly using the SIFT algorithm,resulting in a low tolerance of registration and even a failure in proper registration.After the image enhancement,the feature points extracted will appear inaccurate.In order to obtain higher registration accuracy,a registration algorithm based on mutual information is added after SIFT algorithm registration.2.An improved mutual information registration algorithm is proposed.The traditional mutual information algorithm only considers the grayscale statistical information of the image and ignores the correlation between image pixels.In order to overcome this shortcoming,an image gradient angle mutual information registration algorithm is adopted,which effectively improves the registration accuracy.3.A weighted average method for dynamically merging regions is proposed.Existing methods can not handle the ghosting and exposure differences in the image splicing process well.To solve this problem,an image stitching algorithm is proposed.First,find the best sutures in the images to be stitched using dynamic programming methods.Then find the dynamic fusion region near the fusion line for weighted fusion.Using multiple sets of images as data for registration and splicing experiments,the effectiveness of the above algorithm was verified.